AI-powered diagnostic and operations platform that reasons across your entire infrastructure stack.
MEHO (Machine Enhanced Human Operator) connects to your infrastructure -- Kubernetes, VMware, Prometheus, Loki, ArgoCD, Jira, and more -- and uses AI agents to diagnose problems across system boundaries in a single conversation.
Ask MEHO why your pods are crashing and it will trace from application logs through service mesh to Kubernetes events, node resources, underlying VM metrics, and the Git commit that caused it -- connecting the dots you'd normally spend hours on.
Think Claude Code, but instead of reasoning across files in a codebase, MEHO reasons across systems in an infrastructure.
Prerequisites: Docker and Docker Compose, plus an LLM API key (Anthropic recommended; OpenAI and Ollama also supported).
git clone https://github.com/evoila/meho.git
cd meho
cp env.example .env # Add your LLM API key
docker compose upOpen http://localhost:5173 and start investigating.
Default credentials: admin / admin (change immediately in production).
All 15 connectors are included in the open-source edition.
| Stack | Connectors |
|---|---|
| Infrastructure | Kubernetes, VMware vSphere, Proxmox VE, Google Cloud Platform |
| Observability | Prometheus, Loki, Tempo, Alertmanager |
| CI/CD | ArgoCD, GitHub (repos, PRs, Actions, deployments) |
| Collaboration | Jira, Confluence, Email (IMAP/SMTP) |
| Generic | REST (any OpenAPI spec), SOAP (any WSDL) |
Each connector provides typed operations with trust classification (READ/WRITE/DESTRUCTIVE), topology entity extraction, and natural-language query support.
| Layer | Technology |
|---|---|
| Backend | Python 3.13+, FastAPI, SQLAlchemy, Alembic |
| Frontend | React 19, TypeScript, Vite, TailwindCSS, Zustand v5 |
| AI Agents | PydanticAI with ReAct loop, cross-system investigation skills |
| Database | PostgreSQL 15 with pgvector for hybrid search |
| Cache | Redis |
| Auth | Keycloak OIDC with JWT validation |
| Topology | React Flow + elkjs layout engine |
| Embeddings | Local TEI (bge-m3) by default, Voyage AI optional |
| Storage | MinIO (S3-compatible) for document uploads |
MEHO supports multiple LLM providers. Set MEHO_LLM_PROVIDER in your .env file:
| Provider | Config | Notes |
|---|---|---|
| Anthropic Claude (recommended) | ANTHROPIC_API_KEY=sk-ant-... |
Best reasoning quality. Opus 4.6 for investigation, Sonnet 4.6 for utility tasks. |
| OpenAI | OPENAI_API_KEY=sk-... |
GPT-4o. Good alternative. |
| Ollama | OLLAMA_BASE_URL=http://host.docker.internal:11434 |
Fully local. No API keys needed. See Ollama setup. |
See env.example for the full configuration reference.
- Cross-system investigation -- trace problems across infrastructure, observability, CI/CD, and collaboration tools in one conversation
- Intelligent data pipeline -- JSONFlux normalizes large API responses into Arrow tables, caches as Parquet, uses DuckDB SQL reduction to keep LLM context manageable
- Topology auto-discovery -- builds a connected infrastructure graph with cross-system entity resolution via SAME_AS edges
- Trust model -- three-tier operation classification (READ/WRITE/DESTRUCTIVE) with approval workflows and audit trail
- Dual-mode chat -- Ask mode for knowledge base Q&A, Agent mode for cross-system investigation
- Three-tier knowledge -- global, connector-type, and connector-instance scoped knowledge with hybrid search (BM25 + semantic) and reranking
- Local embeddings -- TEI sidecar with bge-m3 runs locally by default. No cloud embedding service required.
| Community (open-source) | Enterprise | |
|---|---|---|
| Connectors | All 15 | All 15 |
| Investigation | Full cross-system reasoning | Full cross-system reasoning |
| Topology | Auto-discovery + entity resolution | Auto-discovery + entity resolution |
| Knowledge | Hybrid search + local embeddings | Hybrid search + managed embeddings |
| Auth | Single-user (Keycloak) | Multi-tenant SSO/SAML |
| Audit | -- | Export + compliance reporting |
| Collaboration | -- | Group sessions, team features |
| Support | Community | Priority SLA |
The community edition is fully functional for individual operators. Enterprise adds organizational features: multi-tenancy, corporate SSO, audit compliance, and team collaboration.
Learn more at https://meho.ai.
Full documentation is available at https://docs.meho.ai.
To preview locally:
pip install mkdocs-material neoteroi-mkdocs
mkdocs serveThen visit http://localhost:8001.
We welcome contributions! See CONTRIBUTING.md for guidelines on reporting issues, submitting changes, code style, and our Contributor License Agreement.
AGPL-3.0-only. Copyright (c) 2026 evoila Group.